Master Thesis
Marc Jerónimo Pérez y Ropero
Background
In my Internship I studied the current GRUD, particularly Mg, P and K
Fertilizer requirement models imply \(Y\sim STP + Clay\) & \(P-\text{Export}\sim STP + Clay\)
Currently only stationery measurement of STP are considered
Could a desorption kinext-model better explain the soil status and yield data?
Experimental Setup
- LTE STYCS, all treatment conditions equal except P-fertilization, which is in 6 Levels, 3 were considered(\(P0\),\(P100\),\(P166\))
- 5 Sites regarded; Cadenazzo, Ellighausen, Rümlang-Altwi, Oensingen, Zürich-Reckenholz
- 5 Sites, 4 blocks per site, 6 Treatment-Levels, 4 Repetitions
- Years 2017-2022 were observed, kinetic data was collected for year 2022 and used to predict 2017-2022
Kinetic Model
Flossmann & Richter conducted in 1982 experiments that should: - improve the classification of P-supply in soils - work in tandem with a current STP-method e.g. CAL or Olsen
The net-desorption was observed, a kinetic of first order was assumed: \[\frac{dP}{dt}=k\times(P^S-P)\] When solved with \(P(0)=0\), the following equation was obtained: \[P(t)=P^S\times(1-e^{-kt})\] The researchers estimated \(P^S=P_\text{CAL/Olsen}-P_{H_2O}\) and linearized as follows: \[log(1-\frac{P(t)}{P^S})=-kt\] \(PS\), \(k\) and \(k*PS\) were extracted, \(k*PS\) being the average net-release speed.
Kinetic-Experiment Setup
Variable Introduction
Soil Variables:
- \(P-CO_2\) & \(P-AAE10\) stand for the GRUD STP-measurements in [\(g~P/kg ~ Soil\)]
- \(k\)(\(s^{-1}\)) can be interpreted as the relative speed of net-desorption of orthophosphate
- \(k*PS\)(\(g~Ps^{-1}\)) can be interpreted as the average net-release speed
- \(PS\)(\(mg~P/L~H_2O\)) is the equilibrium concentration of \(PO_4^{3-}\) of the net-desorption experiment
- From the 0-20cm Horizon: Clay-, Silt-,\(C_{org}\)-content and pH
Yield Variables:
- For a year \(X\) and crop \(C\) \(Y_{main-rel}\) stands for \(Y_{main-rel}:=Y_C^{X}/mean(Y_C~\text{in year}~X~\text{in CH})\)
- For every year:site:crop combination the yield was normalised using: \(Y_\text{norm}:=Y/median(Y_{P166})\)
- The P-Export was calculated as the P-Uptake of the main product
- The P-Balance was calculated as the difference \(P_{Fertilized}-\text{P-Export}\)
Research Questions
- I: Is the method presented by Flossmann and Richter (1982) with the double extraction replicable with the soils from the STYCS-trial?
- II: How do GRUD-measurements of STP correlate to the soil properties \(C_\text{org}\)-content, clay-content, silt-content and pH?
- III: Are the kinetic coefficients \(k\) and \(PS\) correlated to soil properties?
- IV: How well can current GRUD methods of STP (\(P-CO_2\) & \(P-AAE10\)) predict the Yield-parameters, P-Export and P-Balance?
- V: How well can the kinetic parameters \(k\) & \(PS\) improve the predict Yield-parameters, P-Export and P-Balance?
Research Question I
I: Is the method presented by Flossmann and Richter (1982) with the double extraction replicable with the soils from the STYCS-trial?
Question II & III
II: How do GRUD-measurements of STP correlate to the soil properties \(C_\text{org}\)-content, clay-content, silt-content and pH? III: Are the kinetic coefficients \(k\) and \(PS\) correlated to soil properties?
The following random structure was chosen:
(1|year) + (1|Site) + (1|Site:block) + (Treatment|Site)
Do P-CO2, P_AAE10, k and PS correlate with soil characteristics?
Coefficient Table for Soil Covariates. Significant codes: 0 '***' 0.001 '**' 0.01 '*' 0.05
| (Intercept) |
***-4.271 |
-0.159 |
***-6.881 |
0.049 |
-1.092 |
-0.043 |
***124.095 |
| k |
|
|
|
|
|
|
72.766 |
| k:log(PS) |
|
|
|
|
|
|
36.002 |
| log(PS) |
|
|
|
|
|
|
1.355 |
| soil_0_20_clay |
0.012 |
-0.004 |
-0.007 |
* 0.012 |
-0.016 |
0.010 |
|
| soil_0_20_Corg |
***0.567 |
-0.003 |
0.216 |
0.062 |
* 0.490 |
* 0.379 |
|
| soil_0_20_pH_H2O |
-0.003 |
** 0.035 |
0.174 |
** -0.086 |
0.067 |
* 0.148 |
|
| soil_0_20_silt |
-0.021 |
** 0.007 |
0.013 |
* -0.019 |
-0.036 |
0.015 |
|
| TreatmentP100 |
***1.063 |
0.002 |
** 1.056 |
***0.195 |
** 0.703 |
***0.744 |
|
| TreatmentP166 |
***1.841 |
-0.031 |
***1.633 |
***0.412 |
***1.271 |
***1.157 |
|
| R2m |
0.858 |
0.399 |
0.666 |
0.678 |
0.661 |
0.519 |
0.029 |
| R2c |
0.950 |
0.697 |
0.912 |
0.833 |
0.782 |
0.905 |
0.800 |
Question IV & V
IV: How well can current GRUD methods of STP (\(P-CO_2\) & \(P-AAE10\)) predict the Yield-parameters, P-Export and P-Balance? V: Can the kinetic parameters \(k\) & \(PS\) predict the Yield-parameters, P-Export and P-Balance?
Yield model summary:
Coefficient Table for Yield Variables. Significant codes: 0 '***' 0.001 '**' 0.01 '*' 0.05
| (Intercept) |
***1.038 |
***0.463 |
***0.970 |
***0.881 |
***98.786 |
***67.321 |
***70.659 |
***124.095 |
| k |
|
|
|
* 1.558 |
|
|
|
72.766 |
| k:log(PS) |
|
|
|
** 0.680 |
|
|
|
36.002 |
| log(PS) |
|
|
|
-0.024 |
|
|
|
1.355 |
| log(soil_0_20_P_AAE10) |
|
***0.140 |
0.020 |
|
|
** 8.788 |
8.099 |
|
| log(soil_0_20_P_CO2) |
***0.163 |
|
0.095 |
|
6.190 |
|
1.155 |
|
| log(soil_0_20_P_CO2):log(soil_0_20_P_AAE10) |
|
|
0.021 |
|
|
|
|
|
| TreatmentP100 |
|
|
|
|
6.546 |
4.352 |
4.062 |
|
| TreatmentP166 |
|
|
|
|
4.838 |
2.232 |
1.599 |
|
| R2m |
0.226 |
0.203 |
0.224 |
0.194 |
0.080 |
0.098 |
0.098 |
0.029 |
| R2c |
0.432 |
0.445 |
0.439 |
0.409 |
0.576 |
0.582 |
0.582 |
0.800 |
P-Export model summary:
Coefficient Table for P-export. Significant codes: 0 '***' 0.001 '**' 0.01 '*' 0.05
| (Intercept) |
***24.761 |
9.594 |
** 24.908 |
***36.785 |
| k |
|
|
|
23.494 |
| k:log(PS) |
|
|
|
10.824 |
| log(PS) |
|
|
|
2.380 |
| log(soil_0_20_P_AAE10) |
|
3.385 |
-0.043 |
|
| log(soil_0_20_P_CO2) |
** 4.757 |
|
* 4.783 |
|
| TreatmentP100 |
2.085 |
3.022 |
2.097 |
|
| TreatmentP166 |
1.150 |
3.338 |
1.166 |
|
| R2m |
0.076 |
0.074 |
0.076 |
0.039 |
| R2c |
0.577 |
0.542 |
0.577 |
0.751 |
P-balance model summary:
Coefficient Table for P-balance. Significant codes: 0 '***' 0.001 '**' 0.01 '*' 0.05
| (Intercept) |
***-24.532 |
-8.080 |
* -24.460 |
***45.816 |
| k |
|
|
|
81.169 |
| k:log(PS) |
|
|
|
31.798 |
| log(PS) |
|
|
|
***19.365 |
| log(soil_0_20_P_AAE10) |
|
-3.680 |
-0.022 |
|
| log(soil_0_20_P_CO2) |
* -5.148 |
|
-5.138 |
|
| TreatmentP100 |
***30.965 |
***29.964 |
***30.974 |
|
| TreatmentP166 |
***54.052 |
***51.704 |
***54.066 |
|
| R2m |
0.628 |
0.641 |
0.627 |
0.521 |
| R2c |
0.808 |
0.798 |
0.807 |
0.784 |
Key Findings
- The estimation \(PS=P_{Olsen}-P_{H_2O}\) did not deliver reasonable and significant models
- P-CO2 did not correlate with clay-content
- k does not correlate with Treatment but with pH and silt-content
- PS was the covariate best predicted by soil properties: \(R^2_m=0.858\)
- \(k*log(PS)\) showed the strongest effect in the prediction of Ynorm
- P-AAE10 did show a string and significant effect in prediction of Yrel
- P-CO2 did show strong effects in both predicting Pexport and Pbalance (however negative in Pbalance)
- \(PS\) showed the strongest effect in predicting P_balance